刚刚，ChatGPT免费版史诗升级！GPT-5.6可以无限白嫖了
Ranking
No observed public metrics; popularity remains neutral/archived.
Merged summary
TL;DR - OpenAI shipped a Chat-focused update to ChatGPT: free users get GPT-5.6 Luna as the default model with unlimited chat messages, while Plus/Pro users get a factuality-tuned GPT-5.6 Sol plus a Codex-style reasoning-effort slider. It signals a deliberate pivot back toward conversational quality after months of industry focus on coding.
- Free tier default moves to GPT-5.6 Luna with no chat message cap (rolling out over the week); image generation and file upload limits remain. Luna also gets a dedicated "thinking" button for harder queries.
- GPT-5.6 Sol was retuned for chat: more focused answers, response length adapted to question complexity, less unnecessary formatting, and willingness to correct the user rather than agree. The update applies only to ChatGPT chat — not the Sol builds used in Work or Codex.
- Internal evals on finance, medical, and legal factuality prompts reported ~62% fewer responses containing at least one factual error for Luna vs. GPT-5.5 Instant, and ~68% fewer for Sol.
- Plus/Pro get a sliding effort control (web, mobile, desktop) that jointly sets model and reasoning effort, so one model drives both instant answers and deep reasoning. Article frames this against ByteDance's Zhang Yiming reportedly urging Seed not to over-index on coding.
Sources (1)
刚刚，ChatGPT免费版史诗升级！GPT-5.6可以无限白嫖了
TL;DR - OpenAI shipped a Chat-focused update to ChatGPT: free users get GPT-5.6 Luna as the default model with unlimited chat messages, while Plus/Pro users get a factuality-tuned GPT-5.6 Sol plus a Codex-style reasoning-effort slider. It signals a deliberate pivot back toward conversational quality after months of industry focus on coding.
- Free tier default moves to GPT-5.6 Luna with no chat message cap (rolling out over the week); image generation and file upload limits remain. Luna also gets a dedicated "thinking" button for harder queries.
- GPT-5.6 Sol was retuned for chat: more focused answers, response length adapted to question complexity, less unnecessary formatting, and willingness to correct the user rather than agree. The update applies only to ChatGPT chat — not the Sol builds used in Work or Codex.
- Internal evals on finance, medical, and legal factuality prompts reported ~62% fewer responses containing at least one factual error for Luna vs. GPT-5.5 Instant, and ~68% fewer for Sol.
- Plus/Pro get a sliding effort control (web, mobile, desktop) that jointly sets model and reasoning effort, so one model drives both instant answers and deep reasoning. Article frames this against ByteDance's Zhang Yiming reportedly urging Seed not to over-index on coding.